Jump to text
On this page
Copy and source
Published source confirmed at last check

Source changed 2026-11-18 21:11:39 UTC · snapshot created 2026-11-18 21:12:35 UTC · last check 2026-11-18 22:05:04 UTC

A.6:4 - Solution — A stack + a classification matrix

A.6:4.1 - Why “stack”: what is stacked, and what “higher and lower” means

The Signature Stack orders reusable boundary declarations and the more use-specific claims that depend on them. Its purpose is to allow a realization or evidence arrangement to change while preserving the declarations and constraints that its use still requires. “Higher” and “lower” name this relation:

  • Higher in the stack = closer to public, reusable boundary intent.
  • Lower in the stack = closer to execution, implementation, and evidence (what is done and observed).

The first four positions below classify boundary claims; the final two distinguish their actual subjects and publication uses:

  1. Signature layer (L: laws or definitions). U.Signature provides the stable declarative boundary: Vocabulary + Laws + Applicability, without runtime gate predicates.

  2. Admission layer (A: entry and application conditions). State when a named action or application is admitted under its governing rule. Mechanism applications use U.Mechanism.AdmissibilityConditions; A.21 gates use the applicable profile rule. U.Mechanism specializes U.Signature through the operation declarations, LawSet, AdmissibilityConditions and Applicability governed by A.6.1. Its admission predicates remain declaration content. Evidence-interface declarations and transport details keep their own claim classification; use A.10 when a replayable source-to-use account is needed, and name carrier-producing Work only when that occurrence is claimed.

    Audit and AssuranceLane: a boundary’s local Audit and observability section states its evidence-interface declarations: carrier classes and required fields, correlation keys, and exposure interface. Mechanism.AuditObservability below is a local publication-section locator, not an A.6.1 content component.

    Prescriptions or claims about individual duties concerning retention, access or enforcement are D-claims. A general prescription remains a claim-bearing episteme; a claim about one obtaining individual duty cites the exact A.2.8 U.Commitment, its actual bearer, and its direct predicate. A system-role kind or assignment may be an applicability ground but is neither bearer nor commitment.

    An MVPK AssuranceLane is a publication face for auditors that explains how to adjudicate the evidence interface. Under CC-A.6.6, the AssuranceLane face references those evidence-interface declarations and relevant claim IDs or canonical locations; its explanation preserves their semantics.

  3. Deontic layer (D: duties, commitments, and grants). Put here a general prescription or a claim about an exact individual duty, recommendation-as-duty, prohibition, commitment, or A6-AW-NORM-GRANT. For an individual duty, cite the exact A.2.8 U.Commitment, actual bearer, constitutive rule, required instituting basis, and direct predicate. Test any responsibility claim separately through its domain predicate or return the missing governor. Other A6-AW-* claims keep their own placement. Reference related L-*, A-*, or E-* claims by ID or canonical location rather than duplicating their constraints.

  4. Observable-effects and evidence layer (E: Work-Effects & Evidence). Each E claim names the actual occurrence or evaluated finding under its subject pattern and the observation/evidence basis needed by its receiving use. The subject may be Work under the conditions below, or a natural, spontaneous or formal change. When the publication uses separate sections, place the claim in Evidence-and-carriers; it may also be presented in an AssuranceLane.

  5. Actual occurrences and realizations (outside the description stack). Substitutable realizations are exercised through dated Work only when each actual performer has its A.13 core and A.15.1 independently admits the occurrence. Add F.6 only when the receiving description also consumes precise assignment-bound attribution through the same obtaining A.13 assignment; missing or failed F.6 alone does not overturn independently admitted Work. If the attribution finding questions a premise consumed by Work admission, recheck that premise and the dependent claims.

    Work may participate in change, production, speech-act effect, evaluation, or evidence production, but each relation or claim must be established under the pattern that defines or constrains it. A.3 and A.3.4 also admit natural, spontaneous, and formal transformations without a performer, assignment, Method, or Work occurrence.

  6. Publication faces. MVPK selects exact epistemes and publication forms for audience-specific face uses. A selected episteme has U.View membership only when E.17.0 conformance to the exact viewpoint episteme obtains; any A.6.3 source-to-receiving construction remains separate. The face designator, publication occurrence, form, rendering, and carrier are not the U.View.

When specifying evidence carriers and correlation rules, traces/spans, logs/log records and metric time series provide examples of carrier schemas and join keys. Choose the forms needed by the evidence claim.

A.6:4.1.1 - AssuranceLane skeleton (informative)

An MVPK AssuranceLane is a publication face that teaches a specific audience how to adjudicate E-* claims against the relevant evidence carriers, including those produced in Work. It cites the boundary’s evidence-interface declarations and explains them without changing their semantics.

Minimal content (suggested):

  • Scope: boundaryRef, version; viewRef and viewpointRef when view or viewpoint identity matters.
  • Carrier inventory: carrier-class and carrier-schema refs (A.7 Carrier) + where to obtain them.
  • E‑claim map: a table keyed by E-* ID with: measurement conditions, carrierRef(s), join and correlation keys, and a reference to the canonical E-* text that defines pass or fail criteria.
  • Operational policies: references to relevant D-* duties (retention, access control, exposure), without redefining them.
  • Limitations: sampling, redaction, missing signals, expected false negatives and false positives.

An AssuranceLane explains adjudication. Add a new boundary claim to its governing source before using it in that face; a permission-looking claim needs its selected A6-AW-* object and basis.

Example (conceptual; uses view/viewpoint identity and the additional hypothetical header case in §4.2):

AssuranceLane:
  viewRef: <ViewId>
  viewpointRef: <ViewpointId>
  boundaryRef: <BoundaryId>
  version: <SemVer or revision>
  evidence:
    - E: E-OBS-1
      carrierRefs: [Carrier.AuthorizationRecord, Carrier.AuditLogEntry]
      measurement:
        conditions: "on every request lacking header X (A-AC-1)"
        vantage: "Operator and auditor pipeline"
        correlation: ["traceId", "requestId"]
      adjudication:
        check: "query audit stream for code=NotAdmissible and join to traceId"
        criteriaRef: "E-OBS-1 (pass or fail criteria live canonically in the E-claim)"
      references: [A-AC-1, D-RET-1, Mechanism.AuditObservability]

The absence condition also covers a request that was not rejected. A claim of complete coverage requires the relevant request population to be accounted for separately from individual query joins.

Default placements (quadrant → stack layer / section):

  • L → Signature.Laws (and, where appropriate, mechanism‑local semantic laws; never runtime gates)
  • A → the canonical source of the governing admission rule; use Mechanism.AdmissibilityConditions for mechanism applications and the applicable profile rule for A.21 gates
  • D → generic prescriptions, individual duties or commitments, recommendations-as-duty, prohibitions, and A6-AW-NORM-GRANT claims at their A.2.8 or A.2.8.PER subject pattern
  • E → actual occurrences, evaluated findings, and evidence claims, including A6-AW-EXERCISE, A6-AW-WEAK, A6-AW-CONFLICT, and A6-AW-SOURCE when those claims are current

Related subject rules (informative):

  • A.6.1 ↔ A‑quadrant: U.Mechanism.AdmissibilityConditions is the canonical source for A-* admission predicates of mechanism applications.
  • A.10 / B.3 ↔ E‑quadrant: when an E-* claim needs a replayable source-to-use account, recover its A.10 evidence-provenance path and bounded use. A missing required path narrows or blocks only the unsupported use. Open B.3 only for a named assurance claim about an exact target and assurance use.
  • A.2.3 and F.12 ↔ D/E separation: a U.PromiseContent states the promised outcome and acceptance criteria; F.12 evaluates fulfilment using Work evidence. A general duty remains normative content, while an obtaining individual duty is one A.2.8 U.Commitment borne by an actual System or other admitted party. Any system-role kind or assignment used to establish applicability stays separate. D-* claims reference A-* and E-* claims by ID or canonical location when needed.

This placement supports change through the dependencies:

  • A realization, audit format or transport mechanism can change while the boundary declarations remain valid, provided their applicable constraints are still satisfied.
  • A changed declaration requires reconsidering the uses that depend on it and communicating any consequential compatibility or version change.

A.6:4.2 - Boundary Discipline Matrix: classify by A.6.B (the Boundary Norm Square)

A.6.B defines the matrix and cross-quadrant reference rules. Use the following summary to classify a claim and the examples to distinguish a source clause from added case assumptions.

The matrix distinguishes four jobs in a boundary account: defining meaning, declaring admission conditions, stating normative content or a deontic position, and reporting occurrences or evidence. Classify the claim by that job, then recover the conditions that settle it under A.6.B.

A declared admission predicate and its evaluation have separate jobs: the predicate is A; a particular evaluation result is E. D includes both generic normative content and a claim about an individually obtaining duty or grant, with different grounds for each. These distinctions let the same subject appear in several connected claims. The selected A6-AW-* row supplies the corresponding permission-word distinction. Actual evaluation or observation does not require inventing Work for a non-Work change.

Operational summary (quadrant → canonical claim layer in the stack):

  • L (Laws & Definitions) → Signature.Laws (truth‑conditional semantics, in‑description)
  • A (Admissibility & Gates) → the canonical source of the governing admission rule: Mechanism.AdmissibilityConditions for mechanism applications; the applicable profile rule for A.21 gates. A predicate may consume an exact grant or finding selected by A.6.B:8.4.1, but it neither creates nor resolves that object.
  • D (Deontics) → generic-prescription or individual-duty A.2.8 claims and A6-AW-NORM-GRANT
  • E (Work-Effects & Evidence) → actual-occurrence, evaluated-finding, and evidence claims, including the applicable E-side A6-AW-* row

Atomicity rule:

If a sentence mixes logical jobs, for example “MUST” plus a gate predicate plus an effect claim, it is not classifiable as a single statement. Per A.6.B, split it into atomic claims so each one has exactly one quadrant and, ideally, an identifier you can reference.

Micro‑template: Atomize → Classify → Place → Identify EntityOfConcern → Register when useful

  1. Split when needed: separate claims that perform different logical jobs.
  2. Assign each claim to exactly one quadrant (L/A/D/E) using the matrix.
  3. Place each claim in its canonical claim layer; use a separate section or publication form when the receiving use needs it.
  4. Anchor A.7: name what each claim is about, separately from the episteme carrying it. Add publication and carrier relations when they change interpretation. For permission-looking wording, bind the direct object and participants required by the selected A6-AW-* row; the selected subject pattern or kind of direct object never supplies the quadrant.
  5. Register when useful: add the atomic claim to the Claim Register if used. Downstream faces cite the claim by ID or canonical location and preserve its meaning in any explanation.

If a Work or reliance claim lacks its required source, narrow it to what the available source supports or hold the dependent use. A local or reversible use may proceed on its own adequate grounds once its stopping condition is stated.

Informative example: mixed → atomic.

Before (mixed, not classifiable yet): “Clients MUST include header X; otherwise the request is invalid and the system logs NotAdmissible.”

Recovered source clauses:

  • “Clients MUST include header X.”
  • “If a request lacks header X, the request is invalid.”
  • “If a request lacks header X, the system logs NotAdmissible.”

Recover the intended invalidity/admissibility meaning before fully classifying the second clause. The third clause states a generic logging rule; it reports no particular observed request.

Additional hypothetical illustration. Suppose a separate boundary policy defines invalidity here as failure of the entry condition below and adds an implementer duty to the quoted Clients duty. Also suppose the observation stated in E-OBS-1 occurred in this hypothetical case:

  • A-AC-1 (Quadrant A, Mechanism.AdmissibilityConditions): hasHeader(req, "X") is a necessary entry condition.
  • D-CL-1 (Quadrant D, Norms-and-commitments): “Client implementers MUST include header X in each request to this boundary (A-AC-1).”
  • E-OBS-1 (Quadrant E, Evidence-and-carriers): “For the selected request req lacking header X (A-AC-1), the system logged NotAdmissible; the observer recovered its AuditLogEntry{code="NotAdmissible"} in the audit stream.” The carrier schema is an additional illustrative choice. Logging depends on the absence of X, including when the request was not rejected.

Informative example: guarantee, SLA, measurement and enforcement.

Before (mixed contract prose): “The service guarantees 99.9% availability per calendar month and MUST keep p95 latency under 200ms; breaches are penalized; operators SHALL alert on violations.”

Recovered source clauses:

  • “The service guarantees 99.9% availability per calendar month.”
  • “The service MUST keep p95 latency under 200ms.”
  • “Breaches are penalized.”
  • “Operators SHALL alert on violations.”

Recover the guarantee’s meaning and bearer, the measurement and acceptance basis, and the breach trigger, penalty and applicable parties. Use A.6.C for the unresolved contract meanings and A.6.B to classify the resulting atoms. The split alone does not settle them.

Additional hypothetical illustration. Suppose a separate policy identifies Provider and the service being measured, states the exclusions and workload W, defines the availability and latency metrics, and sets the two criteria evaluated below. In addition to the quoted alert duty, it requires paging within 5 minutes. The following E-claims assume that the stated evaluations and observation occurred in this hypothetical case; the alert observation concerns a separate violation case.

  • D-SLA-1 (Quadrant D, Commitments and SLA): “Provider SHALL meet the availability and latency criteria evaluated by E-SLA-AVAIL-1 and E-SLA-LAT-1 under the stated exclusions.”
  • E-SLA-AVAIL-1 (Quadrant E, Evidence-and-carriers): “The evaluation of the observed calendar month T reported availability ≥ 0.999, with measurements recorded in carrier UptimeProbeSeries from viewpoint VP.ExternalMonitor.”
  • E-SLA-LAT-1 (Quadrant E, Evidence-and-carriers): “The evaluation under workload W reported latency_p95 < 200ms, with measurements recorded in carrier LatencyMetricSeries from viewpoint VP.Client.”
  • D-OPS-ALERT-1 (Quadrant D, Ops duty): “Operators MUST page on breach of the criteria evaluated by E-SLA-AVAIL-1 or E-SLA-LAT-1 within 5 minutes (additional policy).”
  • E-ALERT-1 (Quadrant E, Evidence-and-carriers): “In the separate violation case, the operator’s page was observed in carrier AlertEvent{ruleId,firedAt,target} and can be joined via incidentId.”

These added policy and observation premises do not resolve the original guarantee, duty bearer or penalty clause.

See A.6.B:4–A.6.B:6 for the normative square, quadrant form constraints, and explicit cross‑quadrant link patterns (notably: D→A, E→A, D→E, and A/E→L).

A.6:4.2.1 - Authority-wording split examples

These examples are informative. They separate authority wording from the evidence, assurance, commitment, gate-passage, or Work claim being made.

Before (mixed): “This API is approved for production use and guarantees safe rollback.”

Recovered source clauses:

  • “This API is approved for production use.”
  • “This API guarantees safe rollback.”

Recover the approval’s direct object and ground under the applicable A6-AW-* row, and the rollback subject and safety predicate. Both source claims remain unresolved until that meaning is supplied.

Additional hypothetical illustration. Assume supplied signature vocabulary defines the API operation and rollback terms; a separate boundary policy gives the request-admission predicate, policy window and exclusions. For E-API-1, additionally assume that a named evaluation of an independently identified rollback occurrence reported success under a stated success criterion and observation basis:

  • L-API-1 (Quadrant L): the API operation and rollback terms are defined in the supplied signature vocabulary.
  • A-API-1 (Quadrant A): a request is admissible only under the named subject, action, object, context, and policy-version predicate.
  • D-API-1 (Quadrant D): the provider policy prescribes maintaining or enforcing A-API-1 under the named window and exclusions. If the claim is instead that one actual provider or operator bears this duty, cite its separately instituted A.2.8 commitment.
  • E-API-1 (Quadrant E): the named evaluation reported rollback success under its stated success criterion. Possible evidence inputs include the named work traces, audit records, or metrics. A gate decision carrier may support the gate-passage claim; rollback execution needs its own occurrence and evidence basis.

In this additional case, A-API-1 applies A6-AW-GATE. The observation of the approval badge is an A6-AW-SOURCE claim; any further approval claim needs separate classification and grounds. The original production approval and safe-rollback guarantee remain unresolved. A success result alone does not establish the unspecified safety claim.

For a filled grant/exercise/evidence case and its near-misses, use A.6.B:8.4.5.4. It applies A6-AW-NORM-GRANT, A6-AW-EXERCISE, and the separate A.10 evidence claim by value.

Then:

  • if appearance hides the prerequisite for action or reliance, enter A.15.4; use A.15 for enactment alignment;
  • if evidence, currentness, or provenance is live, attach the A.10 evidence relation;
  • if a named assurance claim is current, use B.3 for its exact target claim, argument, bounded assurance use and AssuranceResult; otherwise keep trust, readiness, compliance or release questions with their direct patterns;
  • if an actual gate decision or passage is asserted, classify it as a separate E claim and cite the exact A.21 GateDecisionResult, bounded action, applicable GateProfile application, complete required GateCheckApplicationResult set, decisionValue, action consequence, scope/window, and recheck condition; use a short GateCheckRef only for a selected publication structure and a DecisionLog only when audit or reuse is current;
  • if a flow witness or constraint witness is asserted, cite A.20 ConstraintValidity status or witness;
  • if a permission-looking claim is asserted, use the selected A6-AW-* row and its subject pattern; an entry predicate or GateDecisionResult does not substitute for another row;
  • if release, deployment, rollback, or execution Work is asserted, cite the exact A.15.1 dated occurrence; then use only the applicable A.15.1:4.6 row for an application result, A.15.PROD production branch, delivery/transfer relation, evaluation/acceptance relation, or A.10 evidence path. None is an intrinsic Work field;
  • if the phrase is only an action invitation or cue, keep it in A.6.A, A.16, or A.16.1 according to the current kind.

Policy engines, credentials, registers, provenance, and attestations can supply policy decisions, source claims, currentness, or evidence. Classify the observation of a visible permit, badge or registry value under A6-AW-SOURCE. Classify any further claim by what it asserts and check the required grounds for its named direct object and participants.

A.6:4.3 - View membership needs exact viewpoint conformance

MultiViewDescribing makes the candidate episteme and exact viewpoint episteme explicit. The candidate has U.View membership only when E.17.0 conformance obtains. A projection or query may participate in an A.6.3 construction, but that construction does not establish membership. MVPK separately uses publication face designators (PlainView, TechCard, InteropCard, AssuranceLane) and their E.17 profiles. E.17:5.2 specifies the declared publication-face kind values.

The stack therefore requires:

  • Every published face use identifies the selected episteme and its separate reader/use declaration. Name the exact viewpoint episteme through U.ViewpointRef when U.View membership or viewpoint identity is used; name the publication occurrence, form, and carrier when those identities change publication or reliance. The face designator is not any of those objects.
  • Calling the selected episteme a U.View requires E.17.0 conformance; a face label, viewpoint reference, projection history, or publication does not establish it.
  • Per E.17 (“no new semantics”), a face MUST NOT introduce a new semantic commitment or any new object or claim selected through A6-AW-*. A face MAY add informative explanation, examples, and cross-references that preserve the source claims. Normative face prose cites the canonical L/A/D/E claim ID or location and direct object; it may faithfully paraphrase the claim. Add any new boundary claim to its canonical source before publishing it on a face. Use verbatim text when exactness is critical or disputed.
  • Per E.17 and publication-face and publication-form discipline (face‑kind closure), a publication package that claims MVPK alignment MUST NOT introduce additional MVPK face kinds (e.g., “EvidenceCard”, “NormsCard”) as if they were first‑class kinds; if you need local headings, keep them as sections within the selected faces.

A.6:4.4 - Recover the separate claims in contract wording

When “the API contract” leaves a consequential ambiguity, use A.6.C to ask only the live questions: what was promised, what was said or instituted, what governance position obtains, and what happened. Use A.15.1:4.6 to separate any dated Work from the result, production, delivery/transfer, evidence, or acceptance claims made. Clear wording, including a semantic guarantee or recoverable ordinary metonymy, needs no unpacking record.

  • Promise content (promise content; U.PromiseContent, A.2.3): what is promised to be made available to eligible consumers — a promise, not execution (U.Work).
  • Utterance and publication: identify what is stated and published in the boundary descriptions or faces. Identify any actual speech-act Work under A.2.9 when that occurrence or its instituting effect is claimed; publication alone does not establish such an act.
  • Commitment (individual deontic relation; U.Commitment, A.2.8): whether one actual admitted System or other party is obligated, recommended-as-duty, or prohibited from doing something under an exact constitutive rule and required instituting basis. A system-role kind or assignment may help satisfy that rule’s applicability conditions; neither is the duty bearer or the commitment relation. A commitment does not establish responsibility, which needs its own direct domain predicate or a missing-governor result.
  • Permission-looking claim: do not make Permission a bundle part or quadrant. Select one A6-AW-* row for each atomic claim and cite its direct object.
  • Performed Work (A.15.1): identify the dated occurrence, performer, Method, time and containing System using the conditions in 4.1. Precise assignment-bound attribution is the separate F.6 claim when needed; its absence or failure does not revoke the Work.
  • Result or consequence (A.15.1:4.6 dispatch): only when current, name the exact A.6.1 application/result binding or subject-specific WorkResultRelation, A.15.PROD production branch, A.3.4 change, evaluation result, delivery/transfer relation, or acceptance relation.
  • Evidence (A.10): only when a receiving use relies on one of those claims, name the claim-bound evidence path and carrier.

In A.6 terms:

  • The signature supplies the boundary’s declared vocabulary, laws and applicability; it is not itself a promiser or obligor (A.7).
  • Deontic claims use A.2.8 for generic prescriptions or separately obtaining individual duties and commitments, and A6-AW-NORM-GRANT for the current norm/grant branch. Other permission-looking claims keep the placement and object named by their selected row.
  • Classify each atomic operational “guarantee” claim as L (truth-conditional law), A (entry predicate), D (generic prescription, individual commitment, or current grant), or E (actual exercise, evaluated result, work effect, or measured property with evidence).

Optional-object example. SVC-DEPLOY-1 states promise content. Admitted system ReleaseManager-4 performs SA-4711 : U.SpeechAct under ReleaseManager-4@ReleaseShift; the exact policy may institute COM-4711 : U.Commitment or PER-4711 : GrantedPermissionRelation@Context. Later admitted system Operator-7 performs DeployRun-4711 : U.Work under its covering assignment.

If the application returns ReleaseArtifact-4711, cite the exact A.6.1 result binding or an already governed WorkResultRelation; if that artifact is delivered, cite a separately obtaining transfer relation defined by its subject pattern; if acceptance is claimed, cite the criterion, evaluation result, and acceptance relation; identify dated evaluation Work only when the claim depends on that occurrence.

An A.10 path may support whichever one of those claims is relied on. Omit every absent object: the Work can occur without a result, delivery, acceptance, or evidence-use claim.

Use A.6.C — Contract Unpacking for Boundaries for the expanded account and the same A.15.1:4.6 dispatch.

A.6:4.5 - Where statements go (classification examples)

Informative. Classification examples for learning the discipline; they do not add requirements beyond A.6:7.

The table below uses near‑everyday spec phrases. The same visible words appear in different quadrants depending on what they do.

The A.7 primary layer field below identifies the claim’s EntityOfConcern, with its kind when needed; it is not a layer selected from the quadrant. The concern may itself be a Description episteme or publication carrier and remains separate from the episteme carrying the claim.

IDExample statement (typical wording)Matrix quadrantPut it under…A.7 primary layer
L-1“op f is defined iff P(x) holds.”LSignature → Laws (Definition:)Operation f; the relation of x to f is unresolved
L-2“For all requests, idempotencyKey is unique per subject.”LSignature → Laws (Invariant:)Requests’ idempotencyKey values; the request population and subject are unresolved
A-1“The mechanism may be applied only if tokenValid.” (rewrite as predicate: admissible(req) implies tokenValid(req))AMechanism → AdmissibilityConditions (entry gate)Entry predicate admissible(req) implies tokenValid(req)
A-2“A request is admissible only if header X is present.”AMechanism → AdmissibilityConditions (mechanism entry case)Request-entry predicate requiring header X
D-1“Client implementers MUST satisfy A-2.”DNorms-and-commitments: a general prescription unless one exact A.2.8 individual commitment and actual bearer are also identified; reference the gate by ID or canonical locationNormative rule requiring client implementers to satisfy A-2
D-2“Authors MUST publish a versioned MVPK face for this boundary.”DConformance Checklist and publication norms (authoring plane)Normative rule requiring authors to publish this boundary’s versioned MVPK face
D-3“Operators SHOULD rotate keys every 90 days.”DNorms: state the prescription; if an individual duty is claimed, identify its actual bearer, direct A.2.8 predicate, and any separately obtaining system-role assignment used only for applicabilityNormative rule recommending that operators rotate keys every 90 days
D-4“Implementers MUST expose audit‑log carriers via endpoint /audit.”DNorms-and-commitments (exposure duty) about carriersNormative rule requiring implementers to expose audit-log carriers via /audit
D-5“The vendor commits to 99.9% availability over window T (SLA).”DCommitments and SLA: identify the actual admitted vendor System or other A.2.8 party as duty bearer, the direct commitment predicate, constitutive rule, required basis, window, and exclusions; any system-role assignment is only a possible applicability groundThe vendor’s individual availability commitment
E-1“LedgerBalance-L17 changed from 80 to 65 across interval T under the stated account-continuity rule.”EA.3/A.3.4 actual transformation claim; no Work is inferred from the delta aloneChange of LedgerBalance-L17 across T
E-1-EVID“AuditRecord-L17 evidences E-1 for audit use under the stated source, window, and A.10 path.”EEvidence relation and carrier for the already named changeA.10 support relation from AuditRecord-L17 to claim E-1 for audit use
D-6“Operators MUST retain audit‑log carriers for 30 days.”DRetention policy (deontic) about carriersNormative rule requiring operators to retain audit-log carriers for 30 days
E-2“latency_p95 ≤ 200ms under workload W using measurements recorded in carrier LatencyMetricSeries from collector C.”EMeasured-property claim with measurement conditions; subject unresolvedEntity whose latency is measured: unresolved; LatencyMetricSeries is the measurement carrier, not that missing referent

Notes:

  • The classification is not just about modal verbs. “Shall” can be D (a duty) or A (a gate behavior). “Guarantees” can be D (a commitment) or E (a measured property). The matrix forces disambiguation.
  • If a sentence combines a duty with an entry condition, split it into (A) a gate predicate (A-*) and (D) either a general prescription or a claim about one exact U.Commitment borne by an actual System or other admitted party (D-* referencing the gate by ID or canonical location). When observability matters, add an E claim only on a separate actual observation or result basis. A requirement to produce or retain logs is D; an expectation or plan supplies no actual E result. A system-role kind or assignment may establish applicability only through an independently obtaining rule; neither bears the duty.
  • When something needs to be enforceable but is mathematical, prefer predicate blocks rather than deontic language in the L/A blocks, per E.8’s deontics vs admissibility guidance.

A.6:4.6 - Classification sanity rules (informative, concept-level)

These are writing diagnostics, not tool requirements.

  • An RFC keyword is used as an obligation operator inside a Definition, invariant or admissibility predicate → separate the D obligation from the L/A claim; write the definition or entry condition as a predicate.
  • E-* with no exact actual occurrence or evaluated predicate, or with a carrier but no evidence relation for the claimed use → incomplete effect/evidence claim. Ground Work through A.15.1 only when it obtains; otherwise use A.3/A.3.4 or the exact interaction or causal-use pattern. A carrier supports the claim but does not create the effect.
  • D-* that re-states an A-*/L-* predicate instead of citing its ID or canonical location → drift risk (prefer “MUST satisfy A-…”).
  • A face introduces new L/A/D/E content not present in the canonical claim set → view-fork. Recover the direct object and classify the new claim—duty/commitment/grant in D; exercise/evaluated finding/evidence in E; gate in A—then add it to its canonical source before face publication. Informative commentary may explain existing claims without adding boundary semantics.
  • “The system or service SHALL …” where the phrase does not name a direct behavior claim, general prescription, or exact individual commitment with its actual bearer and constitutive basis → unresolved subject and modality. Recover the System or other party, state an actual E-* behavior claim separately only when its basis is supplied, and state either the normative content or the direct A.2.8 commitment. A service label, system-role kind, or assignment proves none of these claims.

A.6:4.7 - Use the claim dependencies to judge a boundary change

Start with the receiving use: which clients, operations and observations must continue to work, under which conditions? Read the existing boundary claims and the proposed change against that use. The stack makes the different constraints findable; compatibility depends on their content and connections.

  1. Recover what the receiver relies on. Include the meaning of inputs and results, admitted inputs, relevant interaction order and state changes. Add a duty, grant or evidence requirement when this use depends on it. Preserve the conditions and exceptions attached to each claim.
  2. Identify what changes. Distinguish changing a definition, an admission condition, normative content, an implementation or an observation method. A change can affect several of these. Use the claim references to follow the consequences to the receiving use, including an A predicate that consumes a current D grant or E finding under A.6.B:6.4.
  3. Compare the old and proposed behavior in that use. For a replacement offered to existing clients, determine whether previously admitted inputs remain admitted and whether the results, interactions and state changes still satisfy the old claims. When representations differ, state the correspondence used for this comparison. A stronger entry condition can exclude an old valid call; a weaker result condition can remove a property its client needs. Preserving a protocol or mutable state also requires its own sequence or history conditions.
  4. Establish the affected claims on their own grounds. A declaration comparison can establish a conditional compatibility argument. Claims about an actual implementation, obtained grant, observed result or retained evidence need their corresponding grounds. A changed telemetry format can leave the operation semantics intact while breaking the auditor’s ability to check a result.
  5. Choose and communicate the continuation. Preserve the old boundary, adapt the receiver or supply an adapter with a justified correspondence, or expose a changed contract and its affected uses. Reuse unaffected claims and evidence while their conditions hold. Keep an unresolved compatibility premise explicit.

Use the smallest description that carries these dependencies. For one operation whose meaning and use are already clear, its existing preconditions, postconditions and invariants can supply the comparison. A protocol specification can carry the interaction constraints; a policy and its evaluation interface can carry an authorization decision. Add the L/A/D/E organization where connected semantic, admission, governance and evidence claims otherwise obscure what a change affects. The labels organize those claims; the relevant contract, protocol, policy or subject Method supplies the preservation test.